Context

Overview

Context is everything Polylane knows about your system. Connect your clouds, code, and observability providers, and the platform builds a live context graph that every agent reasons over.

Context is the first step of the Polylane loop. Before the platform can detect, investigate, or remediate anything, it needs a live model of your system. That model comes from what you connect.

What you connect

  • Integrations: your code host, chat, and observability providers, plus any Model Context Protocol server. Integrations register tools agents can call, and provider events like github.push or vercel.deployment become automation triggers.
  • Clouds: your cloud providers. Connecting a cloud account syncs its resources into the context graph and keeps them in sync.
  • Repositories: the repositories Polylane indexes so agents can search and read your code.

What Polylane builds from it

  • Topology: the graph of your resources and the relationships between them. Every synced resource becomes a node; edges capture how traffic, deploys, and dependencies flow. The galaxy view renders the same graph spatially.
  • Memories: durable knowledge about your system. Investigation and automation runs pull relevant memories into the agent's context, and agents write back what they learn, so an investigation in July benefits from what an agent found in June.
flowchart LR
  C[Clouds] --> G[Context graph]
  R[Repositories] --> G
  I[Integrations] --> G
  G --> T[Topology]
  G --> M[Memories]
  T --> A[Agents]
  M --> A
  A --> D[Detection and investigation]

How agents use context

Every downstream step reads from the context graph:

  • Detection evaluates the resources you connected and raises issues. Change intelligence compares each sync against the last one to record what changed. With nothing connected, there is nothing to check.
  • Investigation agents walk the topology to scope blast radius, query your observability integrations for evidence, and read repositories to connect symptoms to code.
  • Remediation agents call the tools your integrations expose. An autofix needs a connected repository to open a pull request against.

Connect one cloud and one observability provider first. Add your code host when you want autofixes or code-aware investigations.

Next steps